A Method for Searching and Locating Concealed Beam Radiation Sources on the Sea Surface Based on Amphibious Aircraft

Abstract

This invention belongs to the field of wireless communication technology and relates to a method for searching and locating concealed beam radiation sources on the sea surface based on amphibious aircraft. By periodically sampling radio signals using the aircraft's onboard antenna, the detection probability and false alarm probability of high signal-to-noise ratio signals are obtained. An energy distribution probability map, characterizing the uplink beam energy distribution of the radiation source, is constructed and updated based on Bayesian criteria. The amphibious aircraft then schedules its search area according to this map: if the probability of an uplink beam energy distribution in the current local search area is low, the aircraft uses a virtual force-guided Levy flight path planning to jump out of the current local search area and fly to the nearest less searched local search area; otherwise, the actor-evaluator flight path planning based on a kernel function remains in the current local search area for further searching. This search area scheduling mechanism improves the search efficiency of amphibious aircraft in locating concealed beam radiation sources on the sea surface.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, and in particular relates to an autonomous search and positioning method for a concealed beam radiation source on the sea surface based on an amphibious aircraft. Background Technology

[0002] Amphibious aircraft possess superior amphibious capabilities that land-based aircraft and ships lack, and their application in maritime reconnaissance has gained increasing attention in recent years. Currently, marine reconnaissance instruments that collect ocean information have a significant impact on maritime safety. These instruments communicate with satellites via onboard communication modules to transmit their findings. This type of marine reconnaissance has already posed a major threat to maritime security. However, for aircraft conducting reconnaissance over the ocean, the underwater portion of these instruments is often small, almost invisible compared to the vast ocean, making them difficult to identify using optical methods; or they may be disguised, making them difficult for humans to distinguish as reconnaissance instruments.

[0003] When a marine reconnaissance instrument's communication module communicates with a satellite, it can be considered a concealed beam radiation source on the sea surface. Using radio reconnaissance to identify and locate such sources is a viable approach. To enable rapid search and location of concealed beam radiation sources over sea areas using amphibious aircraft, an efficient and accurate search and location method, including trajectory planning and positioning algorithms, is crucial. Summary of the Invention

[0004] The purpose of this invention is to propose a method for searching and locating concealed beam radiation sources on the sea surface based on amphibious aircraft, which conducts radio reconnaissance on the satellite communication modules of marine reconnaissance instruments to achieve the purpose of locating the marine reconnaissance instruments.

[0005] The technical solution of the present invention:

[0006] A method for searching and locating concealed beam radiation sources on the sea surface based on amphibious aircraft includes:

[0007] Step 1: Obtain the path point position of the amphibious aircraft in the current flight iteration n, and confirm the aircraft's grid position in the energy distribution probability map;

[0008] Step 2: Determine the flight decision for flight iteration n based on the search area scheduling mechanism. The flight decision includes virtual force-guided Levy flight trajectory planning and kernel function-based actor-evaluator trajectory planning.

[0009] Step 3: Execute flight decision for flight iteration n and determine the expected path point location for the next flight iteration n+1. During flight, the amphibious aircraft uses its onboard antenna to periodically sample radio signals and determine in real time whether they are valid positioning signals, while simultaneously updating the energy distribution map in real time.

[0010] Step 4: When planning the Levi flight path guided by virtual force, determine whether the amphibious aircraft has sampled a valid positioning signal during flight.

[0011] Step 5: If no valid positioning signal is sampled, fly to the expected path point of flight iteration n+1 and return to execute the operation of step 1;

[0012] Step 6: If a valid positioning signal is sampled, the expected path point position of flight iteration n+1 is discarded, and the position of the first sampled valid positioning signal is used as the path point position of flight iteration n+1. The flight decision of flight iteration n+1 is transformed into actor-evaluator trajectory planning based on kernel function.

[0013] Step 7: When the flight decision is based on the actor-evaluator trajectory planning of the kernel function, the amphibious aircraft directly flies to the expected path point position of flight iteration n+1 and records the effective positioning signals and sampling positions sampled during the flight.

[0014] Step 8: Use a positioning algorithm to process the accumulated valid positioning signals and sampling locations to obtain the estimated location of the radiation source;

[0015] Step 9: Determine whether the estimated location of the radiation source meets the conditions for accurate positioning;

[0016] Step 10: If the accurate positioning conditions are met, the flight mission terminates and the estimated location of the radiation source is output.

[0017] Step 11: If the accurate positioning conditions are not met, return to Step 1.

[0018] Furthermore, the path point location is the initial stop location for each flight iteration, which is also the location where the flight decision for this flight iteration is made.

[0019] Furthermore, the energy distribution map is defined as follows: a square grid of the same size is defined to cover the airspace plane above the search sea area, and each grid is assigned a certain energy distribution probability value, which represents the energy distribution probability of the uplink beam of the radiation source projected onto the grid.

[0020] Furthermore, the search area scheduling mechanism is as follows: if the average energy distribution probability of the square grid where the aircraft is located and the neighboring square grids is less than the set scheduling threshold, the flight decision for this flight iteration is virtual force-guided Levy flight trajectory planning; otherwise, the flight decision is kernel function-based actor-evaluator trajectory planning.

[0021] Furthermore, the virtual force-guided Levi flight trajectory planning is as follows: an artificial potential field is constructed using an energy distribution map, the aircraft is subjected to virtual gravity generated by each square grid, and the virtual gravity dominates the aircraft's flight direction; the aircraft's flight step length is determined by the Levi flight with a minimum step length restriction.

[0022] Furthermore, the kernel function-based actor-evaluator trajectory planning is as follows: In the actor-evaluator reinforcement learning method, the actor is composed of a policy function, which generates actual executable trajectory planning actions according to the policy based on the trajectory planning decision state observed by the aircraft; the evaluator evaluates the quality of the actions taken by the actor through the state value function or the action value function, and uses time difference error to guide the update of the policy function and the value function.

[0023] Furthermore, a Gaussian kernel function is used to approximate the policy function and the state value function. The inputs to both the policy function and the state value function are the flight path planning decision states of the aircraft. The output of the state value function is the value corresponding to the flight path planning decision state, and the output of the policy function is the probability corresponding to each flight path planning action in the flight path planning decision state.

[0024] Furthermore, the effective positioning signal, i.e. the signal sampled by the airborne antenna unit, has an average estimated received signal power greater than the set effective positioning signal power threshold.

[0025] Furthermore, the energy distribution map is updated in real time, that is, the probability of detection and the probability of false alarm of the sampled signal obtained by the aircraft being a valid positioning signal are updated according to the Bayesian criterion to update the energy distribution probability value of the square grid in the energy distribution map.

[0026] Furthermore, the positioning algorithm is as follows: the pitch angle and azimuth angle of the effective positioning signal are measured using the MUSIC angle of arrival measurement algorithm, and the estimated radiation source position is obtained by combining the aircraft position and flight altitude through geometric relationships; considering that the single estimated radiation source position is not accurate enough, the cluster center estimation method is used to determine the cluster center of multiple estimated radiation source positions to obtain a more accurate estimated radiation source position.

[0027] The positioning condition is that the estimated radiation source location data contained in the circle centered on the cluster center is greater than the set maximum estimated data threshold for the cluster center.

[0028] The beneficial effects of this invention are as follows: The method of this invention utilizes an aircraft-borne antenna to periodically sample radio signals, obtaining the detection probability and false alarm probability of high signal-to-noise ratio signals. Based on Bayesian criteria, an energy distribution probability map characterizing the uplink beam energy distribution of the radiation source is constructed and updated. The search area scheduling of the amphibious aircraft is then performed based on this energy distribution probability map: if the probability of an uplink beam energy distribution of the radiation source in the current local search area is low, the aircraft will jump out of the current local search area through virtual force-guided Levy flight trajectory planning and fly to the least searched local search area near its current location; otherwise, it will remain in the current local search area for further searching through kernel function-based actor-judge trajectory planning. This search area scheduling mechanism effectively improves the search efficiency of amphibious aircraft in locating concealed beam radiation sources on the sea surface. Attached Figure Description

[0029] Figure 1 This is a flowchart of the autonomous search and positioning method for a concealed beam radiation source on the sea surface based on an amphibious aircraft, provided by the present invention.

[0030] Figure 2 This invention provides a map of the scene and energy distribution of amphibious aircraft searching for concealed beam radiation sources on the sea surface.

[0031] Figure 3 This is a data processing diagram of radiation source estimation location provided by the present invention. Detailed Implementation

[0032] The following description of embodiments provides a more detailed explanation of the specific implementation of the present invention, including the shape and structure of each component, the relative positions and connections between the parts, the function and working principle of each part, the manufacturing process, and the operation and use methods, in order to help those skilled in the art to have a more complete, accurate, and in-depth understanding of the concept and technical solution of the present invention.

[0033] The specific steps of the method for searching and locating concealed beam radiation sources on the sea surface based on amphibious aircraft provided by this invention are as follows:

[0034] Step 1: Obtain the path point position of the amphibious aircraft in the current flight iteration n, and confirm the aircraft's grid position in the energy distribution probability map;

[0035] In this step:

[0036] like Figure 2 As shown, the aircraft flies at a relatively constant altitude above the search area. This information can be obtained from a radio altimeter;

[0037] Compared to the vast search area, the aircraft can be considered a physical point, and its waypoint location... It is the initial stop position for each flight iteration, which can be obtained by fusing information from the satellite positioning system and the inertial navigation system.

[0038] Define N x N y A grid of identical squares covers the airspace plane (aircraft flight plane) above the search area. Each grid is assigned a certain probability value for the uplink beam energy distribution of the radiation source. , , This allows the construction of an energy distribution probability map. By calculating the information entropy of the energy distribution probability within the grid, the energy distribution uncertainty of the grid can be obtained.

[0039] .

[0040] Step 2: Determine the flight decision for flight iteration n based on the search area scheduling mechanism. The flight decision includes virtual force-guided Levy flight trajectory planning and kernel function-based actor-evaluator trajectory planning.

[0041] In this step:

[0042] The search area scheduling mechanism is as follows:

[0043] Average energy distribution probability of the grid where the aircraft is located and its neighboring grids Set scheduling threshold At that time, the flight decision selection was based on the Levi flight trajectory planning guided by virtual forces;

[0044] Otherwise, the flight decision is selected as the actor-evaluator trajectory planning based on the kernel function.

[0045] The virtual force-guided flight path planning for Levi is as follows:

[0046] Its step size is given by the following formula:

[0047] ;

[0048] in, It is the shortest step length of Levi's flight guided by the virtual force set up. These are the parameters that control the Lévy flight step size; Lévy flight parameters. Levi's flight parameters ,in, , , It is the Gamma function. It is the Lévy flight scale parameter that adjusts the ratio of the long step length and the short step length of Lévy flight.

[0049] Its direction is determined by the sum of the virtual gravitational forces acting on the aircraft:

[0050] ;

[0051] in, N represents the airspace plane covering the search area. x N y A set of grids; It's the virtual gravity that the grid exerts on the aircraft:

[0052] ;

[0053] in, It is the current signal sampling time slot or the energy distribution probability map update time slot; Representing path points To grid center The distance, i.e. ; This indicates the maximum distance at which the virtual gravity generated by the grid can exert its effect. This represents a given distance threshold, if Virtual gravity is maintained at a distance level; It is by point to The unit vector, i.e. , indicating the direction of the virtual gravitational force generated by a single grid; Given an energy distribution uncertainty threshold, the energy distribution uncertainty is greater than... The grid creates a virtual gravitational pull on the aircraft; Represents a zero vector; It is a scale parameter that adjusts the magnitude of virtual gravity; , It is a parameter that adjusts the relationship between virtual gravity and the uncertainty of energy distribution in each grid; It is a parameter that adjusts the relationship between virtual gravity and the distance from the aircraft to the center of each grid.

[0054] Therefore, when planning the flight path for Levi's flight guided by virtual forces, the flight decision is as follows:

[0055] ;

[0056] The kernel-based actor-evaluator trajectory planning is as follows:

[0057] The algorithm flow for actor-evaluator trajectory planning based on kernel functions is as follows: At the beginning of the algorithm, the feature dictionary of the aircraft with respect to the policy function and state value function is initialized. and , corresponding preference function weight vector and state value function weight vector In flight iteration At that time, the aircraft was at the waypoint Observed trajectory planning decision status According to the parameterized strategy function Output the probability corresponding to each action, and select the trajectory planning action to be executed. The aircraft then carried out... And move to the next state. Receive rewards Flight iteration End and reach the waypoint According to the parameterized state value function, and And calculate the time difference error. Use the time difference error to adjust the weights respectively. and The feature is updated, and the decision features are determined based on a feature selection method based on approximate linear correlation detection. and Can they be added to a dictionary? and Until the motion iterations. Search task completed or waypoint The average energy distribution probability of the grid where the user resides is lower than the scheduling threshold. Levi's flight path planning, which involves ending the flight or conducting virtual force guidance.

[0058] Define the trajectory planning and decision state observed by the aircraft. for:

[0059] ;

[0060] in, The aircraft is at the waypoint The average estimated received signal-to-noise ratio of the sampled signal; It is a path point The probability of energy distribution in the grid in which it is located.

[0061] Define the aircraft motion space Each trajectory planning action in the process is as follows:

[0062] ;

[0063] in It is the stride length of the movement. , It is the action space Size, It is the angle between adjacent actions. The aircraft is in the trajectory planning and decision-making state. trajectory planning maneuvers from Selected from the options.

[0064] The parameterized policy function outputs the probability of each action through the SoftMax function:

[0065] ;

[0066] Among them, the preference function of the policy function Approximation using kernel functions , and These represent the weight vector and kernel vector of the preference function, respectively. It is the aircraft's flight iteration Time-Policy Function Feature Dictionary The number of features in the equation. Similarly, the parameterized state value function is also approximated by a kernel function:

[0067] ;

[0068] in, and These are the weight vector and kernel vector of the state-value function, respectively. It is the aircraft's flight iteration Time-state value function feature dictionary The number of features in the dictionary. Using a Gaussian kernel, the aircraft is defined and stored in the feature dictionary of the policy function. and state value function feature dictionary The first in The features are respectively and Therefore, the kernel vector and No. The elements of the row are as follows:

[0069] ;

[0070] ;

[0071] in, and These represent the characteristic length metrics of the trajectory planning decision state and the trajectory planning action, respectively.

[0072] Using time difference error to guide the updates of the policy function and value function includes: outputting the parameterized state-value function. and Calculate the time difference error:

[0073] ;

[0074] in, Discount factor; For aircraft in flight path planning and decision-making state Execute trajectory planning actions The reward obtained subsequently. The weight vector of the preference function of the policy function. The update method is as follows:

[0075] ;

[0076] Weight vector of state value function The update method is as follows:

[0077] .

[0078] Decision features are determined using a feature selection method based on approximate linear correlation detection. or Can they be added to the feature dictionary? and This includes: calculating feature selection error. , where the kernel vector nuclear matrix , It is the feature selection threshold that determines whether a feature can be added to the feature dictionary; if : ;otherwise .

[0079] Step 3: Execute flight decision for flight iteration n and determine the expected path point location for the next flight iteration n+1. During flight, the amphibious aircraft uses its onboard antenna to periodically sample radio signals and determine in real time whether the positioning signal is valid, while simultaneously updating the energy distribution probability map in real time.

[0080] In this step:

[0081] Expected path point location for flight iteration n+1 Add the flight decision vector to the path point positions of flight iteration n.

[0082] When making flight decisions for Levi's flight path planning guided by virtual forces: ;

[0083] When flight decision-making is based on kernel function-based actor-evaluator trajectory planning: .

[0084] Use an airborne antenna periodically (signal sampling period is...) time slot The corresponding time is The process involves sampling radio signals and determining in real time whether a location signal is valid, including: the signal vector received in a single snapshot during airborne flat panel antenna signal sampling is:

[0085] ;

[0086] in, It is the signal vector received by the airborne antenna. This refers to the number of airborne flat panel antenna elements; This indicates that the radiation source emits a signal and satisfies the following conditions: , The emission power of the radiation source; This represents the channel matrix between the aircraft and the radiation source. This refers to the number of flat panel antenna elements of the radiation source; It is additive white Gaussian noise and obeys the following rules: ,in It is the size of The identity matrix, It was identified as white noise power; This is the precoding vector used by the radiation source for beamforming, which determines the direction and power of the radiation source beam. Assume the number of snapshots per signal sampling is... Let the received signal matrix be . The average estimated received signal power of the airborne antenna element is:

[0087] ;

[0088] in, Represents F-norm operations. This is the estimated noise power, obtained by sampling noise multiple times and calculating the average noise power. An effective positioning signal power threshold is set. ,like:

[0089] ;

[0090] The signal sampled at this time is considered a valid positioning signal.

[0091] The energy distribution probability map is updated in real time, including time slots. Airplane in the grid During signal sampling, the grid is updated according to the Bayesian criterion. Energy distribution probability:

[0092] ;

[0093] in, Corresponding time slot The average estimated received signal power of the airborne antenna element; PD,k and P F,k These correspond to the aircraft in the time slots. The acquired sampled signals represent the detection probability and false alarm probability of the effective positioning signal. The detection probability and false alarm probability can be given by the following formulas:

[0094] ;

[0095] ;

[0096] in, It is the complementary cumulative distribution function of the standard normal distribution. ; It is a time slot The average estimated signal-to-noise ratio of the sampled signal.

[0097] Step 4: When planning the Levi flight path guided by virtual force, determine whether the amphibious aircraft has sampled a valid positioning signal during flight.

[0098] Step 5: If no valid positioning signal is sampled, fly to the expected path point of flight iteration n+1 and return to execute the operation of step 1;

[0099] Step 6: If a valid positioning signal is sampled, the expected path point position of flight iteration n+1 is discarded, and the position of the first sampled valid positioning signal is used as the path point position of flight iteration n+1. The flight decision of flight iteration n+1 is transformed into actor-evaluator trajectory planning based on kernel function.

[0100] Step 7: When the flight decision is based on the actor-evaluator trajectory planning of the kernel function, the amphibious aircraft directly flies to the expected path point position of flight iteration n+1 and records the effective positioning signals and sampling positions sampled during the flight.

[0101] Step 8: Use a positioning algorithm to process the accumulated valid positioning signals and sampling locations to obtain the estimated location of the radiation source;

[0102] In this step, the positioning algorithm includes two parts: measuring the angle of arrival of the effective positioning signal and processing the data for estimating the location of the radiation source.

[0103] The classic MUSIC angle-of-arrival (AOA) algorithm is used to measure the elevation and azimuth angles of the effective positioning signal and estimate the location of the radiation source. This includes obtaining the covariance matrix of the sampled signal when the airborne antenna samples the effective positioning signal.

[0104] ;

[0105] Perform eigenvalue decomposition on the covariance matrix of the sampled signal. ,in and They represent the rank as Sum of rank a vector diagonal matrix, The number of incident signals is 1, therefore ; and Let represent the estimated vector matrices for the signal subspace and noise subspace, respectively. The MUSIC algorithm, based on the orthogonality of linear combinations of the received signal direction vector and any number of noise subspace eigenvectors, estimates the angle of arrival by searching for the angle corresponding to the peak of the spatial spectrum function. The array spatial spectrum function can be expressed as:

[0106] ;

[0107] in, This represents the array steering vector of the receiving airborne antenna. The estimated azimuth angle of the corresponding effective positioning signal. and pitch angle for:

[0108] ;

[0109] Assuming the aircraft is at the sampling location Once a valid positioning signal is sampled, the azimuth angle is estimated based on the valid positioning signal. Pitch angle and the aircraft's flight altitude This allows us to obtain an estimated location of the radiation source. for:

[0110] ;

[0111] .

[0112] Processing the estimated location data of the radiation source yields a more accurate location, such as... Figure 3 As shown, this includes the fact that the estimated location of a single radiation source may deviate too much from its true location, making it inaccurate. This occurs in each flight iteration. Path points The accumulated estimated locations of radiation sources are combined into a cluster of estimated locations of radiation sources. ,in, yes The index number of the data in the middle. It is the drone at the waypoint The number of estimated radiation source locations is collected. K-means clustering is the most basic and commonly used clustering algorithm, which searches for the location of radiation sources through iterative clustering. The cluster partitioning scheme minimizes the loss function corresponding to the cluster center (in this invention). For k-means clustering, the clustering loss function can be defined as follows: Location data And accurate estimation of the radiation source location (cluster center). Sum of squared errors:

[0113] ;

[0114] in The solution is the cluster center. for The centroid of the data, i.e.:

[0115] ( );

[0116] However, despite setting an effective positioning signal power threshold, the angle of arrival measurement of some effective signals may still be inaccurate due to noise. This could lead to excessive deviations in the estimated positions of some radiation sources from their actual positions. Therefore, it is necessary to exclude radiation source estimation data with excessive deviations. Thus, the following cluster center estimation algorithm is used to obtain... The cluster center is used as a more accurate radiation source estimation location. The algorithm automatically finds the region with the highest data distribution density for radiation source estimation locations to adapt to the actual distribution of radiation source estimation location data and improve positioning accuracy. The algorithm flow for cluster center estimation of radiation source estimation location data is as follows: Given a minimum data threshold for cluster center estimation... Cluster center estimated location distance threshold and cluster center estimated distance threshold In the 0th algorithm iteration, for any estimated position... Make its radius The estimated locations within this range constitute a new cluster of estimated locations. ,in, yes The amount of data in the dataset. If for all... All satisfied If the radiation source estimation location data processing algorithm is not executed, there will be no output; if any satisfy ,Pick ,remember Iteratively estimate the location of the radiation source for the 0th algorithm iteration. And record for ,remember for And calculate the estimated location of the radiation source in the first algorithm iteration. for Centroid of the data: ( ). In the first In the next algorithm iteration, if the radiation source location is estimated... The estimated location of the radiation source compared to the previous algorithm iteration Not satisfied Then make radius The location data within constitutes a new cluster (in yes (Number of radiation source location data in the data), to calculate the number of radiation source location data. The radiation source location is estimated in each iteration of the algorithm. for Centroid of the data: ( If satisfied); ,remember The algorithm iteration terminates, outputting a more accurate estimated location of the radiation source (cluster center). ( ).

[0117] Step 9: Determine whether the estimated location of the radiation source meets the conditions for accurate positioning;

[0118] In this step, the positioning conditions are: ;in, Given the cluster centers, estimate the maximum data threshold. This is the last iteration of the radiation source estimation location data processing algorithm. Location data volume .

[0119] Step 10: If the accurate positioning conditions are met, the flight mission terminates and the estimated location of the radiation source is output.

[0120] Step 11: If the accurate positioning conditions are not met, return to Step 1.

[0121] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for searching and locating concealed beam radiation sources on the sea surface based on amphibious aircraft, characterized in that, include: Step 1: Obtain the path point position of the amphibious aircraft in the current flight iteration n, and confirm the grid position of the aircraft in the energy distribution probability map; the energy distribution probability map is: defined as square grids of the same size, covering the airspace plane above the search sea area, and each grid is assigned a certain energy distribution probability value, that is, characterizing the energy distribution probability of the uplink beam of the radiation source projected onto the grid. Step 2: Determine the flight decision for the current flight iteration n according to the search area scheduling mechanism. The flight decision includes virtual force-guided Levy flight trajectory planning and kernel function-based actor-evaluator trajectory planning. The search area scheduling mechanism is as follows: if the average energy distribution probability of the square grid where the aircraft is located and the neighboring square grids is less than the set scheduling threshold, the flight decision for this flight iteration is selected as virtual force-guided Levy flight trajectory planning; otherwise, the flight decision is selected as kernel function-based actor-evaluator trajectory planning. The kernel function-based actor-evaluator trajectory planning is as follows: in the actor-evaluator reinforcement learning method, the actor is composed of a policy function. Based on the trajectory planning decision state observed by the aircraft, it generates an actual executable trajectory planning action according to the policy. The evaluator evaluates the quality of the action taken by the actor through the state value function or the action value function, and uses time difference error to guide the update of the policy function and the value function. Step 3: Execute the flight decision for the current flight iteration n and determine the expected path point location for the next flight iteration n+1. During the flight, the amphibious aircraft uses its onboard antenna to periodically sample radio signals and determine in real time whether they are valid positioning signals, while simultaneously updating the energy distribution probability map in real time. Step 4: When planning the Levi flight path guided by virtual force, determine whether the amphibious aircraft has sampled a valid positioning signal during flight. Step 5: If no valid positioning signal is sampled, fly to the expected path point of flight iteration n+1 and return to execute the operation of step 1; Step 6: If a valid positioning signal is sampled, the expected path point position of flight iteration n+1 is discarded, and the position of the first sampled valid positioning signal is used as the path point position of flight iteration n+1. The flight decision of flight iteration n+1 is transformed into actor-evaluator trajectory planning based on kernel function. Step 7: When the flight decision is based on the actor-evaluator trajectory planning of the kernel function, the amphibious aircraft directly flies to the expected path point position of flight iteration n+1 and records the effective positioning signals and sampling positions sampled during the flight. Step 8: Use a positioning algorithm to process the accumulated valid positioning signals and sampling locations to obtain the estimated location of the radiation source; Step 9: Determine whether the estimated location of the radiation source meets the conditions for accurate positioning; Step 10: If the accurate positioning conditions are met, the flight mission terminates and the estimated location of the radiation source is output. Step 11: If the accurate positioning conditions are not met, return to Step 1.

2. The method according to claim 1, characterized in that, The path point location refers to the initial stop location of each flight iteration, which is also the location where the flight decision for this flight iteration is made.

3. The method according to claim 1, characterized in that, The virtual force-guided Levi flight trajectory planning is as follows: an artificial potential field is constructed using an energy distribution probability map, the aircraft is subjected to virtual gravity generated by each square grid, and the virtual gravity dominates the aircraft's flight direction; the aircraft's flight step length is determined by the Levi flight with a minimum step length restriction.

4. The method according to claim 1, characterized in that, The Gaussian kernel function is used to approximate the policy function and the state value function. The inputs to both the policy function and the state value function are the flight path planning decision states of the aircraft. The output of the state value function is the value corresponding to the flight path planning decision state, and the output of the policy function is the probability corresponding to each flight path planning action in the flight path planning decision state.

5. The method according to claim 1, characterized in that, The effective positioning signal, i.e. the signal sampled by the airborne antenna unit, has an average estimated received signal power greater than the set effective positioning signal power threshold.

6. The method according to claim 1, characterized in that, The real-time updating of the energy distribution probability map involves determining the detection probability and false alarm probability of the sampled signal being a valid positioning signal obtained by the aircraft, and updating the energy distribution probability value of the square grid in the energy distribution probability map according to the Bayesian criterion.

7. The method according to claim 1, characterized in that, The positioning algorithm is as follows: the pitch and azimuth angles of the effective positioning signal are measured using the MUSIC angle of arrival measurement algorithm, and the estimated radiation source position is obtained by combining the aircraft position and flight altitude through geometric relationships; considering that the single estimated radiation source position is not accurate enough, the cluster center estimation method is used to determine the cluster center of multiple estimated radiation source positions to obtain a more accurate estimated radiation source position. The positioning condition is that the estimated radiation source location data contained in the circle centered on the cluster center is greater than the set maximum estimated data threshold for the cluster center.

Images